[llvm] [AMDGPU] Enable vectorization of i8 values. (PR #134934)

[Jeffrey Byrnes via llvm-commits]

================
@@ -0,0 +1,1096 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii -passes=slp-vectorizer %s | FileCheck -check-prefix=GFX7 %s
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -passes=slp-vectorizer %s | FileCheck -check-prefix=GFX8 %s
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -passes=slp-vectorizer %s | FileCheck -check-prefix=GFX9 %s
+
+define protected amdgpu_kernel void @arith_1(<16 x i8> %invec, ptr %out, i32 %flag) {
----------------
jrbyrnes wrote:

I think these arith_* tests are negative tests for enabling i8 vectorization, but this variant should be deleted as there are not multiple chains

https://github.com/llvm/llvm-project/pull/134934